Evans believes that the reason for the failure of any researcher to find evidence of the lives that it was predicted would be saved in the '70s is because of an effect that he calls "selective recruitment". "Selective recruitment" is the observation that "belt wearers are less risky drivers than belt nonwearers." The "process by which it is the safer than average drivers who switch from non-wearing to wearing has been labeled 'selective recruitment'."(Evans p255). Evans has provided a chart (Figure 10.5, p 259) that shows the expected fatalities reductions plotted against belt use rate. The curve shows fatality reductions that appear to vary as a polynomial function of the proportion of belt use rate.
However in the paper by Reinfurt DW, Campbell BJ, Stewart JR Stutts JC. it is obvious that some people, even in the absense of any sanction whatsoever, will incorrectly report that they were wearing a belt when they were not.
Then, in June 1995 a research note by the NHTSA on "Estimating Lives Saved by Restraint Use in Potentially Fatal Crashes" stated that the method used to calculate lives saved had to be revised because overreporting of belt use on police accident reports (self reported) had distorted statistics so that "safety belt use for crash survivors continues to appear unreliably high".
Evans has treated drivers as an inhomogeneous group in deriving the "selective recruitment" theory. He has recognized that there exist groupings of high risk drivers who do not wear seat belts.
Evans believed that it was the failure of the group of "high risktaking" drivers to wear seat belts (while destined by it's dangerous driving habits to have fatal accidents) that was preventing the expected saving of lives from occurring in time series studies.
It is clear that such groups do exist. Drivers who are alcohol affected are a high risk group. Alcohol affected drivers constitute a group that drives between 0.001% and 0.5% of kilometers driven during any week, depending on the definition of "alcohol affected".(See Voas et al).
In 1996 in the USA, 23.8% of fatalities were alcohol affected, and 30.5% of those fatalities were belted. Of the remaining 77.2% of non-alcohol-affected fatalities, 71.1% of drivers wore belts. From this it seems reasonable to deduce that alcohol affected drivers wear belts less than half as frequently as drivers not affected by alcohol, and have fatal accidents at a much greater rate.
However the reliance on self reporting is a fatal flaw in this otherwise interesting study.
"Selective Recruitment" is a special case of "Simpson's Paradox" in statistics.
By treating drivers as an inhomogeneous group in deriving the selective recruitment observation Evans has recognized that there exist groupings of high risk drivers who do not wear seat belts.
Simpson's paradox makes it clear that if 1% of the drivers produce 23.8% of the fatalities, and if that 1% of drivers wears belts less than half as frequently as the remaining 99% of drivers, then disaggregating those numbers is a necessary precondition to making an accurate empirical calculation of the effectiveness of seat belts.
That is what the evidence document does. The fatality rate of
a population with the high risk (selective recruitment) group "alcoholics"
is isolated, and the effectiveness of seat belts in the non-alcohol affected
population is calculated.
return to index page